Etica & Politica / Ethics & Politics, 2003, 1
http://www.units.it/~dipfilo/etica_e_politica/2003_1/7_monographica.htm
Lucas against Mechanism (*)
David Lewis
University of California at Los Angeles
J. R. Lucas argues in Minds, Machines, and Gödel, (1) that his potential output of truths of
arithmetic cannot be duplicated by any Turing machine, and a fortiori cannot be duplicated by
any machine. Given any Turing machine that generates a sequence of truths of arithmetic,
Lucas can produce as true some sentence of arithmetic that the machine will never generate.
Therefore Lucas is no machine
I believe Lucas’s critics have missed something true and important in his argument. I shall
restate the argument in order to show this. Then I shall try to show how we may avoid the antimechanist conclusion of restated argument.
As I read Lucas, he is rightly defending the soundness of a certain infinitary rule of inference.
Let L be some adequate formalisation of the language of arithmetic; henceforth when I speak
of sentences, I mean sentences of L, and when I call them true, I mean that they are true on the
standard interpretation of L. We can define a certain effective function Con from machine
tables to sentences, such that we can prove the following by metalinguistic reasoning about L.
C1. Whenever M specifies a machine whose potential output is a set S of sentences, Con (M) is
true if and only if S is consistent.
C2. Whenever M specifies a machine whose potential output is a set S of true sentences, Con
(M) is true.
C3. Whenever M specifies a machine whose potential output is a set S of sentences including
the Peano axioms, Con (M) is provable form S only if S is consistent.
Indeed, there are many such functions; let Con be any chosen one of them. Call  a
consistency sentence for S if and only if there is some machine table M such that  Con (M)
and S is the potential output of the machine whose table is M. Now I can state the rule R which
I take Lucas to be defending.
R. If S is a set of sentences and  is a consistency sentence for S, infer  from
S.
Lucas’s rule R is a perfectly sound rule of inference: if the premises S are all true, then by C2
so is the conclusion  . To use R is to perform an inference in L, not to ascend to metalinguistic
reasoning about L. (It takes metalinguistic reasoning to show that R is truth-preserving, but it
takes metalinguistic reasoning to show that any rule is truth-preserving.)
Lucas, like the rest of us, begins by accepting the Peano axioms for arithmetic. (Elementary or
higher-order; it will make no difference.) A sentence  is a theorem of Peano's arithmetic if
and only if  belongs to every superset of the axioms which are closed under the ordinary
rules of logical inference. Likewise, let us say that a sentence  is a theorem of Lucas
arithmetic if and only if  belongs to every superset of the axioms which are closed under the
ordinary rules of logical inference and also closed under Lucas’s rule R. We have every bit as
much reason to believe that the theorems of Lucas arithmetic are true as we have to believe that
the theorems of Peano arithmetic are true: we believe the Peano axioms, and the theorems
come from them by demonstrability truth-preserving rules of inference. Knowing this, Lucas
stands ready to produce as true any theorem of Lucas arithmetic. (2)
1
Suppose Lucas arithmetic were potential output of some Turing machine. Then it would have a
consistency sentence  . Since Lucas arithmetic is closed under R,  would be a theorem of
Lucas arithmetic. Then  would, trivially, be provable from Lucas arithmetic. Then, by C3,
Lucas arithmetic would be inconsistent. Lucas arithmetic would contain falsehoods, and so
would the Peano axioms themselves. Therefore, insofar as we trust the Peano axioms, we know
that Lucas arithmetic is not the potential output of any Turing machine. Assuming that any
machine can be simulated by a Turing machine - an assumption that can best be taken as a
partial explication of Lucas’s concept of a machine - we know that neither is it the potential
output of any machine. Thus if Lucas arithmetic is the potential output of Lucas, then Lucas is
no machine.
So far, so good; but there is one more step. Although Lucas has good reason to believe that all
theorems of Lucas arithmetic are true, it does not yet follow that his potential output is the
whole of Lucas arithmetic. He can produce as true any sentence which ha can somehow verify
to be a theorem of Lucas arithmetic. If there are theorems of Lucas arithmetic that Lucas
cannot verify to be such, then, his potential output falls short of Lucas arithmetic. For all we
know, it might be the potential output of a suitable machine. To complete his argument that he
is no machine - at least, as I have restated the argument - Lucas must convince us that he has
the necessary general ability to verify theoremhood in Lucas arithmetic. If he has that
remarkable ability, the he can beat the steam drill - and no wonder. But we are given no reason
to think that he does have it.
It is no use appealing to the fact that we can always verify theoremhood in any ordinary
axiomatic theory - say, Peano's arithmetic - by exhibiting a proof. True, if we waive practical
limitations on endurance; but Lucas arithmetic is not like an ordinary axiomatic theory. Its
theorems do have proofs; but some of these proofs are transfinite sequences of sentences since
Lucas’s rule R can take an infinite set S of premises. These transfinite proofs will not be
discovered by any finite search, and they cannot be exhibited and checked in any ordinary way.
Even the finite proofs in Lucas arithmetic cannot be checked by any mechanical procedure, as
proofs in an ordinary axiomatic theory can be. In order to check whether Lucas’s rule R has
been used correctly, a checking procedure would have to decide whether a given finite set S of
sentences was the output of a machine with a given table M. But a general method for deciding
that could easily be converted into a general method for deciding whether any given Turing
machine will halt on any given input - and that, we know, is impossible.
We do not know how Lucas verifies theoremhood in Lucas arithmetic, so we do not know how
many of its theorems ha can produce as true. He can certainly go beyond Peano arithmetic, and
he is perfectly justified in claiming the right to do so. But he can go beyond Peano's arithmetic
and still be a machine, provided that some sort of limitations on his ability to verify
theoremhood eventually leave him unable to recognise some theorem of Lucas arithmetic, and
hence unwarranted in producing it as true. (3)
Notes
(*) First published in Philosophy, 1969, 44, pp. 231-233. © Cambridge University Press.
Republished by permission. back
(1) Minds, Machines, and Godel, Philosophy, 36 (1961): 112-127. back
(2) Lucas arithmetic belongs to a class of extensions of Peano arithmetic studied by A. M.
Turing in “Systems of Logic based on Ordinals”, Proceedings of the London Mathematical
Society, sec. 2, 45 (1939): 161-228, and by S. Feferman in “Transfinite Recursive Progressions
of Axiomatic Theories”, Journal of Symbolic Logic, 27 (1962): 259-316. back
(3) I am indebted to George Boolos and Wilfrid Hodges for valuable criticisms of an earlier
version of this paper. back
2